Natural gas is one of the harder things to forecast, because it is really several forecasts wearing a trench coat. Demand is mostly a weather story. Supply is a production and LNG story. Price sits on top of storage, which reflects both. Any one of them can move and drag your forecast off course.
Why gas models drift so fast
A heating-season demand model built on last winter assumes last winter patterns. Then a mild stretch arrives, or a cold snap lands early, and the load you forecast is off by enough to matter. On the supply side, LNG export flows, pipeline maintenance, and production changes all shift faster than a model retrained a quarter ago can track. Storage ties it together, so an error in demand or supply flows straight into the price you are trading.
The cost shows up in the bid
When a gas forecast drifts, you feel it in storage decisions, hedges, and bids. A demand number that is a little high means you over-commit. A supply read that is a little stale means you misjudge the balance. None of it announces itself. You just notice the position was not where the forecast said it would be.
Keep the model, fix the drift
You do not need to rebuild your gas models to fix this. A layer sits on top of whatever you run, watches how each forecast performs against what actually happened, learns the patterns it keeps missing, and corrects the output as conditions move. Your models and data stay private.
Weather, storage, LNG, production. You cannot stop the drivers from moving. You can stop your forecast from lagging behind them.
- Demand that tracks the current weather regime, not last season.
- Supply that keeps up with LNG and production shifts.
- Price and storage forecasts that reflect both, continuously.
Weather is where we first proved the core, improving every major model we tested. For gas, where weather is the single biggest driver of demand, that is exactly the input that most needs to stay honest.
Forecasting gas supply, demand, or price?
Talk to us